[daniel-p-carvalho]

Summary

Update CONFIG_MCUBOOT_VERSION to d6267bdb7f60a00f068b750f9efa2b6cb36bf47e, which includes mcu-tools/mcuboot#2908.

The NuttX port of flash_area_get_sector() looked the flash device up by fa_off + off, which only succeeds for off == 0. For any other offset it returned -errno (0) without filling the sector, and MCUboot used uninitialized stack values as the erase offset and size. With MCUBOOT_SWAP_USING_MOVE the loader then erased pages in a loop and left the primary slot corrupted, with no error and no reset. The fix looks the device up by flash area id and returns an offset relative to the area.

Impact

Users of boot/mcuboot get the MCUboot revision with the fixed flash_area_get_sector(). Without it, MCUBOOT_SWAP_USING_MOVE does not work on the NuttX port. No change to Kconfig symbols or the build process; only the default of the version string changes. Users who pin their own MCUBOOT_VERSION are not affected.

Testing

Host: Linux x86_64, arm-none-eabi-gcc 15.2.1.
Target: custom STM32G474RE board (out-of-tree, CONFIG_ARCH_BOARD_CUSTOM), MCUboot loader with MCUBOOT_SWAP_USING_MOVE, 2 x 192 KB slots, based on nuttx 7e652be44d and apps 2d68f38f7.

Before: the loader erased pages in a loop and the primary slot was corrupted (flash_area_erase: ID:1 offset:134271849 length:536870996 in the MCUboot log).

After, without any local patch:

• Loader builds (50488 B) and boots the confirmed v1 image.
• v2 written to the secondary slot: swap completes and v2 runs.
• Reset without confirming: rollback to v1.
• v2 again, then mcuboot_confirm and reset: v2 stays, no new swap.

[cederom]

Thank you @daniel-p-carvalho for fixing the issue directly on the upstream project! :-)

[xiaoxiang781216]

@daniel-p-carvalho need reduce the firmware size:

====================================================================================
Configuration/Tool: same70-qmtech/mcuboot-loader,CONFIG_ARM_TOOLCHAIN_GNU_EABI
2026-10-22 01:53:11
------------------------------------------------------------------------------------
  Cleaning...
  Configuring...
  Disabling CONFIG_ARM_TOOLCHAIN_GNU_EABI
  Enabling CONFIG_ARM_TOOLCHAIN_GNU_EABI
  Building NuttX...
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed

  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0

100 1881k    0 1881k    0     0  2616k      0 --:--:-- --:--:-- --:--:-- 2616k
arm-none-eabi-ld: /github/workspace/sources/nuttx/nuttx section `.ramfunc' will not fit in region `flash'
arm-none-eabi-ld: region `flash' overflowed by 120 bytes
make[1]: *** [Makefile:240: nuttx] Error 1
make: *** [tools/Unix.mk:575: nuttx] Error 2
make: Target 'all' not remade because of errors.

[daniel-p-carvalho]

The overflow comes from the size of the new MCUboot (about 370 bytes more) on a configuration that already uses 99.7% of its flash, not from this change. I sent apache/nuttx#20909, which enables LTO in same70-qmtech:mcuboot-loader as same70-xplained already does, and lists other ways to save flash. Once it is merged this CI check should pass on a re-run.

[daniel-p-carvalho]

@xiaoxiang781216 apache/nuttx#20909 (LTO in same70-qmtech:mcuboot-loader) is now merged, so the flash overflow that failed Linux (arm-07) should be gone. Could you please restart the CI for this PR? Re-running only the failed jobs will not help, since the nuttx tree used by the build is fixed when the run starts. If the new run is green, could you merge it as well? Thanks.